Spatiotemporal models for predicting high pollen concentration level of Corylus, Alnus, and Betula

Identifying influence factors and thresholds of the next day's pollen concentration in different seasons using interpretable machine learning

The prevalence of pollen allergies is a pressing global issue, with projections suggesting that half of the world's population will be affected by 2050 according to the estimation of the World Health Organization (WHO). Accurately forecasting pollen allergy risks requires identifying key factors and their thresholds for aerosol pollen. To address this, we developed a technical framework combining advanced machine learning and SHapley Additive exPlanations (SHAP) technology, focusing on Beijing. By analyzing meteorological data and vegetation phenology, we identified the factors influencing next-day's pollen concentration (NDP) in Beijing and their thresholds. Our results highlight vegetation phenology data from Synthetic Aperture Radar (SAR), temperature, wind speed, and atmospheric pressure as crucial factors in spring. In contrast, the Normalized Difference Vegetation Index (NDVI), air temperature, and wind speed are significant in autumn. Leveraging SHAP technology, we established season-specific thresholds for these factors. Our study not only confirms previous research but also unveils seasonal variations in the relationship between radar-derived vegetation phenology data and NDP. Additionally, we observe seasonal fluctuations in the influence patterns and threshold values of daily air temperatures on NDP. These insights are pivotal for improving pollen concentration prediction accuracy and managing allergic risks effectively.

Development of a Random Forest model for forecasting allergenic pollen in North America

Pollen allergies have negative impacts on health. Information about airborne pollen concentration can improve symptom management by guiding choices affecting timing of medicines and pollen exposure. Observations provide accurate pollen concentrations at point locations. However, in the contiguous United States and southern Canada (CUSSC), observations are sparse, and sampling is often seasonal, intermittent or both. Modeling pollen concentration can fill in the gaps with estimates where direct observations are unavailable and also provide much-needed forecasts. The goal of this study is to develop and evaluate statistical models that predict daily pollen concentrations using a machine learning Random Forest algorithm. To evaluate our methods, we made retrospective forecasts of four pollen types (Quercus, Cupressaceae, Ambrosia and Poaceae), each in one of four CUSSC locations. Meteorological and vegetation conditions were input to the models at city and regional scales. A data augmentation technique was investigated and found to improve model skill. Models were also developed to forecast pollen in locations where there are no observations. Forecast skill in these models were found to be greater than in previous models. Nevertheless, the skill is limited by the spatiotemporal resolution of the pollen observations.

Extension of WRF-Chem for birch pollen modelling-a case study for Poland

In recent years, allergies due to airborne pollen allergens have shown an increasing trend, along with the severity of allergic symptoms in most industrialized countries, while synergism with other common atmospheric pollutants has also been identified as affecting the overall quality of citizenly life. In this study, we propose the state-of-the-art WRF-Chem model, which is a complex Eulerian meteorological model integrated on-line with atmospheric chemistry. We used a combination of the WRF-Chem extended towards birch pollen, and the emission module based on heating degree days, which has not been tested before. The simulations were run for the moderate season in terms of birch pollen concentrations (year 2015) and high season (year 2016) over Central Europe, which were validated against 11 observational stations located in Poland. The results show that there is a big difference in the model's performance for the two modelled years. In general, the model overestimates birch pollen concentrations for the moderate season and highly underestimates birch pollen concentrations for the year 2016. The model was able to predict birch pollen concentrations for first allergy symptoms (above 20 pollen m^-3) as well as for severe symptoms (above 90 pollen m^-3) with probability of detection at 0.78 and 0.68 and success ratio at 0.75 and 0.57, respectively for the year 2015. However, the model failed to reproduce these parameters for the year 2016. The results indicate the potential role of correcting the total seasonal pollen emission in improving the model's performance, especially for specific years in terms of pollen productivity. The application of chemical transport models such as WRF-Chem for pollen modelling provides a great opportunity for simultaneous simulations of chemical air pollution and allergic pollen with one goal, which is a step forward for studying and understanding the co-exposure of these particles in the air.

Understanding hourly patterns of Olea pollen concentrations as tool for the environmental impact assessment

Bioinformatics clustering application for mining of a large set of olive pollen aerobiological data to describe the daily distribution of Olea pollen concentration. The study was performed with hourly pollen concentrations measured during 8 years (2011-2018) in Extremadura (Spain). Olea pollen season by quartiles of the pollen integral in preseason (Q1: 0%-25%), in-season (Q2 and Q3: 25%-75%) and postseason (Q4: 75%-100%). Days with pollen concentrations above 100 grains/m³ were clustered according to the daily distribution of the concentrations. The factors affecting the prevalence of the different clusters were analyzed: distance to olive groves and the moment during the pollen season and the meteorology. During the season, the highest hourly concentrations during the day where between 12:00 and 14:00, while during the preseason the highest hourly concentrations were detected in the afternoon and evening hours. In the postseason the pollen concentrations were more homogeneously distributed during 9-16 h. The representation shows a well-defined hourly pattern during the season, but a more heterogeneous distribution during the preseason and postseason. The cluster dendrogram shows that all the days could be clustered in 6 groups: most of the clusters shows the daily peaks between 11:00 and 15:00 with a smooth curve (Cluster 1 and 3) or with a strong peak (2 and 5). Days included in cluster 9 shows an earlier peak in the morning (before 9:00). On the other hand, cluster 6 shows a peak in the afternoon, after 15:00. Hourly concentrations show a sharper pattern during the season, with the peak during the hours close to the emission. Out of the season, when pollen is expected to come from farther distances, the hourly peak is located later from the emission time of the trees. Significant factors for predicting the hourly pattern were wind speed and direction and the distance to the olive groves.

Geographical Imputation of Missing Poaceae Pollen Data via Convolutional Neural Networks

Airborne pollen monitoring datasets sometimes exhibit gaps, even very long, either because of maintenance or because of a lack of expert personnel. Despite the numerous imputation techniques available, not all of them effectively include the spatial relations of the data since the assumption of missing-at-random is made. However, there are several techniques in geostatistics that overcome this limitation such as the inverse distance weighting and Gaussian processes or kriging. In this paper, a new method is proposed that utilizes convolutional neural networks. This method not only shows a competitive advantage in terms of accuracy when compared to the aforementioned techniques by improving the error by 5% on average, but also reduces execution training times by 90% when compared to a Gaussian process. To show the advantages of the proposal, 10%, 20%, and 30% of the data points are removed in the time series of a Poaceae pollen observation station in the region of Madrid, and the airborne concentrations from the remaining available stations in the network are used to impute the data removed. Even though the improvements in terms of accuracy are not significantly large, even if consistent, the gain in computational time and the flexibility of the proposed convolutional neural network allow field experts to adapt and extend the solution, for instance including meteorological variables, with the potential decrease of the errors reported in this paper.

Applying machine learning to forecast daily Ambrosia pollen using environmental and NEXRAD parameters

Approximately 50 million Americans have allergic diseases. Airborne plant pollen is a significant trigger for several of these allergic diseases. Ambrosia (ragweed) is known for its abundant production of pollen and its potent allergic effect in North America. Hence, estimating and predicting the daily atmospheric concentration of pollen (ragweed pollen in particular) is useful for both people with allergies and for the health professionals who care for them. In this study, we show that a suite of variables including meteorological and land surface parameters, as well as next-generation radar (NEXRAD) measurements together with machine learning can be used to estimate successfully the daily pollen concentration. The supervised machine learning approaches we used included random forests, neural networks, and support vector machines. The performance of the training is independently validated using 10% of the data partitioned using the holdout cross-validation method from the original dataset. The random forests (R= 0.61, R²= 0.37), support vector machines (R= 0.51, R²= 0.26), and neural networks (R= 0.46, R²= 0.21) effectively predicted the daily Ambrosia pollen, where the correlation coefficient (R) and R-squared (R²) values are given in brackets. Three independent approaches-the random forests, correlation coefficients, and interaction information-were employed to rank the relative importance of the available predictors.

Estimating the daily pollen concentration in the atmosphere using machine learning and NEXRAD weather radar data

Millions of people have an allergic reaction to pollen. The impact of pollen allergies is on the rise due to increased pollen levels caused by global warming and the spread of highly invasive weeds. The production, release, and dispersal of pollen depend on the ambient weather conditions. The temperature, rainfall, humidity, cloud cover, and wind are known to affect the amount of pollen in the atmosphere. In the past, various regression techniques have been applied to estimate and forecast the daily pollen concentration in the atmosphere based on the weather conditions. In this research, machine learning methods were applied to the Next Generation Weather Radar (NEXRAD) data to estimate the daily Ambrosia pollen over a 300 km × 300 km region centered on a NEXRAD weather radar. The Neural Network and Random Forest machine learning methods have been employed to develop separate models to estimate Ambrosia pollen over the region. A feasible way of estimating the daily pollen concentration using only the NEXRAD radar data and machine learning methods would lay the foundation to forecast daily pollen at a fine spatial resolution nationally.

Applying Deep Neural Networks and Ensemble Machine Learning Methods to Forecast Airborne Ambrosia Pollen

Allergies to airborne pollen are a significant issue affecting millions of Americans. Consequently, accurately predicting the daily concentration of airborne pollen is of significant public benefit in providing timely alerts. This study presents a method for the robust estimation of the concentration of airborne Ambrosia pollen using a suite of machine learning approaches including deep learning and ensemble learners. Each of these machine learning approaches utilize data from the European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric weather and land surface reanalysis. The machine learning approaches used for developing a suite of empirical models are deep neural networks, extreme gradient boosting, random forests and Bayesian ridge regression methods for developing our predictive model. The training data included twenty-four years of daily pollen concentration measurements together with ECMWF weather and land surface reanalysis data from 1987 to 2011 is used to develop the machine learning predictive models. The last six years of the dataset from 2012 to 2017 is used to independently test the performance of the machine learning models. The correlation coefficients between the estimated and actual pollen abundance for the independent validation datasets for the deep neural networks, random forest, extreme gradient boosting and Bayesian ridge were 0.82, 0.81, 0.81 and 0.75 respectively, showing that machine learning can be used to effectively forecast the concentrations of airborne pollen.

The evaluation of pollen concentrations with statistical and computational methods on rooftop and on ground level in Vienna - How to include daily crowd-sourced symptom data

Background

It is recommended to position pollen monitoring stations on rooftop level to assure a large catchment area and to gain data that are representative for a regional scale. Herein, an investigation of the representativeness of pollen concentrations was performed for 20 pollen types in the pollen seasons 2015-2016 in Vienna for rooftop and ground level and was compared with weather data and for the first time with symptom data.

Methods

The complete data set was analyzed with various statistical methods including Spearmen correlation, ANOVA, Kolmogorov-Smirnov test and logistic regression calculation: Odds ratio and Yule's Q values. Computational intelligence methods, namely Self Organizing Maps (SOMs) were employed that are capable of describing similarities and interdependencies in an effective way taking into account the U-matrix as well. The Random Forest algorithm was selected for modeling symptom data.

Results

The investigation of the representativeness of pollen concentrations on rooftop and ground level concerns the progress of the season, the peak occurrences and absolute quantities. Most taxa examined showed similar patterns (e.g. Betula), while others showed differences in pollen concentrations exposure on different heights (e.g. the Poaceae family). Maximum temperature, mean temperature and humidity showed the highest influence among the weather parameters and daily pollen concentrations for the majority of taxa in both traps.

Conclusion

The rooftop trap was identified as the more adequate one when compared with the local symptom data. Results show that symptom data correlate more with pollen concentrations measured on rooftop than with those measured on ground level.

Predicting the onset of Betula pendula flowering in Poznań (Poland) using remote sensing thermal data

Due to the urban heat island effect, the time of plant pollination might markedly vary within the area of a city. However, existing pollen forecasts do not reflect the spatial variations in the pollen release time within a heterogeneous urban environment. The main objective of this study was to model the spatial pattern of flowering onset (and thus the moment of pollen release) in silver birch (Betula pendula Roth.) in Poznań (Western Poland) using land surface temperature (LST) data and in situ phenological observations. The onset of silver birch flowering was observed at 34 urban and rural sites (973 trees) in Poznań from 2012 to 2014. Forty-four thermal variables were retrieved from MODerate Resolution Imaging Spectroradiometer (MODIS) data. To predict the spatio-temporal distribution of B. pendula flowering onset dates in a city, the ordinary and partial least squares, support vector machine and random forest regression models were applied. The models' performance was examined by an internal repeated k-fold cross-validation and external validation with archival phenological data (2010). Birch flowering began significantly earlier in the urban sites compared to the rural sites (from -1.4 days in 2013, to -4.1 days in 2012). The maximum March LST difference between the urban and rural sites reached 2.4 °C in 2013 and 4.5 °C in 2012. The random forest model performed best at validation stage, i.e. the root mean square error between the predicted and observed onset dates was 1.461 days, and the determination coefficient was 0.829. A calibrated model for predicting the timing of flowering in a heterogeneous city area is an important step in developing a fine-scale forecasting system that can directly estimate pollen exposure in places where allergy sufferers live. Importantly, by incorporating only pre-flowering thermal data into the model, location-specific allergy forecasts can be delivered to the public before the actual flowering time.

Airborne Alternaria and Cladosporium fungal spores in Europe: Forecasting possibilities and relationships with meteorological parameters

Airborne fungal spores are prevalent components of bioaerosols with a large impact on ecology, economy and health. Their major socioeconomic effects could be reduced by accurate and timely prediction of airborne spore concentrations. The main aim of this study was to create and evaluate models of Alternaria and Cladosporium spore concentrations based on data on a continental scale. Additional goals included assessment of the level of generalization of the models spatially and description of the main meteorological factors influencing fungal spore concentrations. Aerobiological monitoring was carried out at 18 sites in six countries across Europe over 3 to 21 years depending on site. Quantile random forest modelling was used to predict spore concentrations. Generalization of the Alternaria and Cladosporium models was tested using (i) one model for all the sites, (ii) models for groups of sites, and (iii) models for individual sites. The study revealed the possibility of reliable prediction of fungal spore levels using gridded meteorological data. The classification models also showed the capacity for providing larger scale predictions of fungal spore concentrations. Regression models were distinctly less accurate than classification models due to several factors, including measurement errors and distinct day-to-day changes of concentrations. Temperature and vapour pressure proved to be the most important variables in the regression and classification models of Alternaria and Cladosporium spore concentrations. Accurate and operational daily-scale predictive models of bioaerosol abundances contribute to the assessment and evaluation of relevant exposure and consequently more timely and efficient management of phytopathogenic and of human allergic diseases.

Predicting the Poaceae pollen season: six month-ahead forecasting and identification of relevant features

In this paper, we approach the problem of predicting the concentrations of Poaceae pollen which define the main pollination season in the city of Madrid. A classification-based approach, based on a computational intelligence model (random forests), is applied to forecast the dates in which risk concentration levels are to be observed. Unlike previous works, the proposal extends the range of forecasting horizons up to 6 months ahead. Furthermore, the proposed model allows to determine the most influential factors for each horizon, making no assumptions about the significance of the weather features. The performace of the proposed model proves it as a successful tool for allergy patients in preventing and minimizing the exposure to risky pollen concentrations and for researchers to gain a deeper insight on the factors driving the pollination season.

Plant Responses to Climate Change: The Case Study of Betulaceae and Poaceae Pollen Seasons (Northern Italy, Vignola, Emilia-Romagna)

Aerobiological data have especially demonstrated that there is correlation between climate warming and the pollination season of plants. This paper focuses on airborne pollen monitoring of Betulaceae and Poaceae, two of the main plant groups with anemophilous pollen and allergenic proprieties in Northern Italy. The aim is to investigate plant responses to temperature variations by considering long-term pollen series. The 15-year aerobiological analysis is reported from the monitoring station of Vignola (located near Modena, in the Emilia-Romagna region) that had operated in the years 1990-2004 with a Hirst spore trap. The Yearly Pollen Index calculated for these two botanical families has shown contrasting trends in pollen production and release. These trends were well identifiable but fairly variable, depending on both meteorological variables and anthropogenic causes. Based on recent reference literature, we considered that some oscillations in pollen concentration could have been a main effect of temperature variability reflecting global warming. The duration of pollen seasons of Betulaceae and Poaceae, depending on the different species included in each family, has not unequivocally been determined. Phenological responses were particularly evident in Alnus and especially in Corylus as a general moving up of the end of pollination. The study shows that these trees can be affected by global warming more than other, more tolerant, plants. The research can be a contribution to the understanding of phenological plant responses to climate change and suggests that alder and hazelnut trees have to be taken into high consideration as sensible markers of plant responses to climate change.